PS 85-208 - Multispectral processing of high resolution satellite imagery to determine the abundance of nesting albatross

Friday, August 11, 2017
Exhibit Hall, Oregon Convention Center
Jane Dolliver1, Robert Suryan1, Anne W. Nolin2, Christopher Noyles3, Ellen Wilt4 and Leah Kenney5, (1)Fisheries and Wildlife, Oregon State University, Corvallis, OR, (2)Oregon State University, (3)U.S. Bureau of Land Management, Anchorage, AK, (4)Bonneville Power Administration, Portland, OR, (5)U.S. Fish and Wildlife Service, Anchorage, AK

All three species of North Pacific albatrosses (Laysan - Phoebastria immutabilis, black-footed- P. nigripes, short-tailed - P. albatrus) are listed as of conservation concern under the IUCN Red List. Counts of albatrosses on breeding colonies provide essential data on annual productivity, attendance, and population trends that supply key metrics to assess population status and effectiveness of management policies or actions. Due to constraints on wildlife monitoring budgets, logistics of accessing colonies, and sensitivities of some breeding areas, we explore the potential for using high resolution multispectral satellite imagery (DigitalGlobe WorldView-2 & 3) to enumerate albatrosses on breeding colonies. Advance image processing techniques include panchromatic-multispectral fusion and linear spectral unmixing. Spectral fusion techniques are designed to enhance identification and enumeration of individual albatrosses. In contract, spectral unmixing techniques separate species-specific spectra from background spectra, identifying the percentage of a pixel that is albatross vs. vegetation or substrate.


We use traditional, colony-based counts in designated plots identifiable by satellite to calibrate satellite-based counts. Preliminary results are promising with performance of either method differing by species and location. In cases where resolution, viewing angle or sun angle limit object-oriented classification of images, spectral unmixing produces a rapid means of detecting albatross in multi-species assemblages, without the knowledge of all spectral endmembers. For surface-nesting seabird colonies these methods could provide an efficient estimate of populations without disturbance to nesting birds or fragile habitats.